From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jeremy Fitzhardinge <jeremy@goop.org>
Subject: Re: [PATCH] Balloon driver crash
Date: Tue, 08 Jun 2010 11:34:42 -0700
Message-ID: <4C0E8D42.1010706@goop.org>
References: <alpine.GSO.2.00.1006031434080.4562@algedi.dur.ac.uk>
	<alpine.LFD.2.00.1006032028110.20613@vega4.dur.ac.uk>
	<1275597402.2782.47.camel@localhost.localdomain>
	<201006031738.15635.dcm@mccr.org>
	<alpine.LFD.2.00.1006072027300.3888@vega3.dur.ac.uk>
	<1275983765.24218.49851.camel@zakaz.uk.xensource.com>
	<alpine.GSO.2.00.1006080927520.13054@algedi.dur.ac.uk>
	<1275986577.24218.49951.camel@zakaz.uk.xensource.com>
	<alpine.GSO.2.00.1006081019000.13054@algedi.dur.ac.uk>
	<4C0E731E.8060308@goop.org>
	<alpine.LFD.2.00.1006081852440.5285@vega2.dur.ac.uk>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Return-path: <xen-devel-bounces@lists.xensource.com>
In-Reply-To: <alpine.LFD.2.00.1006081852440.5285@vega2.dur.ac.uk>
List-Unsubscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xensource.com?subject=unsubscribe>
List-Post: <mailto:xen-devel@lists.xensource.com>
List-Help: <mailto:xen-devel-request@lists.xensource.com?subject=help>
List-Subscribe: <http://lists.xensource.com/mailman/listinfo/xen-devel>,
	<mailto:xen-devel-request@lists.xensource.com?subject=subscribe>
Sender: xen-devel-bounces@lists.xensource.com
Errors-To: xen-devel-bounces@lists.xensource.com
To: M A Young <m.a.young@durham.ac.uk>
Cc: "xen-devel@lists.xensource.com" <xen-devel@lists.xensource.com>, Dave McCracken <dcm@mccr.org>, Ian Campbell <Ian.Campbell@citrix.com>
List-Id: xen-devel@lists.xenproject.org

On 06/08/2010 11:08 AM, M A Young wrote:
> On Tue, 8 Jun 2010, Jeremy Fitzhardinge wrote:
>
>> On 06/08/2010 02:25 AM, M A Young wrote:
>>> This patch fixes an error in commit
>>> f4685d0ed041523d1bd5124c5872459681aca876
>>> "Add hugepage support to balloon driver" that was causing the balloon
>>> driver to crash when shrinking the memory used by Domain-0.
>>
>> Thanks.  But why haven't I seen this?
>
> Good question, but I think you only see the crash in a limited memory
> situation. I don't claim to be an expert in what is going on, but a
> few lines earlier you have the the loop
>
>         for (i = 0; i < nr_pages; i++) {
>                 if ((page = alloc_pages(GFP_BALLOON, balloon_order))
> == NULL) {
>                         nr_pages = i;
>                         need_sleep = 1;
>                         break;
>                 }
>
>                 pfn = page_to_pfn(page);
>                 frame_list[i] = pfn_to_mfn(pfn);
>
>                 scrub_page(page);
>         }
>
> If alloc_pages(GFP_BALLOON, balloon_order) is never NULL during the
> loop then page is left set to be one of the pages you are freeing.
> From then on PageHighMem(pfn_to_page(lpfn)) and PageHighMem(page)
> could easily the same and you will only have a problem if some of the
> pages you are freeing are PageHighMem and some aren't.
>
> If alloc_pages(GFP_BALLOON, balloon_order) is null somewhere in the
> loop (presumably because it can't find enough pages to free) then the
> process will crash.

I think the short answer is that I always run 64-bit dom0, so there's no
highmem...

    J